Localization & State Estimation · ROS 2 / C++ · On-Robot Validation · Learning-Based Control

I build autonomy that runs on real hardware — localization, state estimation and control — and I bring hands-on production-validation rigor from automotive manufacturing to the work.

🤖 Research Assistant, Robust Field Autonomy Lab (Prof. Brendan Englot) @ Stevens — building the localization layer for an autonomous surface vehicle in ROS 2 Jazzy / C++ on an NVIDIA Jetson. AprilTag-based pose estimation with PnP solving, multi-tag quaternion fusion and planar-ambiguity gating: 99.3% localization accuracy, 26 mm radial position error, and >0.5 m outliers cut from 20.3% to 1.2% over a month of on-robot testing and log analysis. Next: a reinforcement-learning controller for station-keeping under varying wave conditions (Fall 2026).

🚗 Before grad school — 3.5 years at AGP eGlass, 2+ of them in design validation and robotic-line work across Tesla, Volvo and GM programs:

• Tesla AP4 Autopilot glass — GD&T validation of camera-bracket position and rotation on the first production runs, Cpk 2.6–8.2 against a ±1.5 mm / 1° tolerance, reported under the Design Validation Process. NMI representative at the Tracy, CA plant (Sep–Dec 2022).
• Volvo V536N / P723N — validated and brought a five-arm ABB robotic line into production in Monterrey, Mexico.
• GM L246 (Cadillac Celestiq) — calibrated the component-placement fixture and devised a ±2σ statistical grouping method that identified the ~46% of each batch needing different settings; adopted by the project team.

🎓 MEng in Robotics @ Stevens Institute of Technology (GPA 3.86, expected Dec 2026) · Graduate Certificate in ML & AI for Mechanical Engineering. Coursework: Autonomous Navigation for Mobile Robots · Modern Control Engineering · Intro to Robotics · Machine Learning in Mech. Eng. · (in progress) Robot Manipulators · Computer Vision · Deep Learning.

🎯 Open to full-time Autonomous Systems / Robotics Software Engineer roles (perception, localization, planning, autonomy) — Texas & California.

Languages

Autonomy & Robotics  Localization & State Estimation · AprilTag Pose Estimation · Multi-Tag Pose Fusion · Motion Planning · Finite-State Machines · PID & Modern Control · Simscape Multibody

Perception & ML  Deep Learning · Reinforcement Learning · Perception Pipelines

Embedded & On-Robot  Real-robot deployment & validation · Sensor calibration · Bag record/replay · Log & telemetry analysis · Industrial robot cells (ABB, KUKA)

Test, Validation & Metrology  Design Validation (DVP) · Process Capability (Cp/Cpk) · GD&T & Dimensional Metrology · Root-Cause Analysis · Engineering Change Requests

AprilTag Localization for an Autonomous Surface Vehicle  ·  ROS 2 Jazzy · C++ · OpenCV  ·  research (RFAL @ Stevens) An absolute-localization layer for an ASV whose primary nav dead-reckons over time. PnP pose solving, transform-chain math, multi-tag quaternion averaging and planar-ambiguity gating on a monocular camera. 99.3% localization accuracy · 26 mm radial error with two tags · >0.5 m outliers reduced 20.3% → 1.2%. Includes the verification procedure and accuracy characterization across the camera field of view.

AprilTag Detector — ROS 2 Camera Front-End  ·  ROS 2 · C++ · Python  ·  research (RFAL @ Stevens) Camera bring-up and detection front-end for the localization stack: gscam integration, per-resolution metric calibration, verification tooling, and bag record/replay so field runs can be reproduced and debugged off the vehicle.

Multi-Robot Warehouse Simulation  ·  MATLAB · Simulink · Simscape Multibody  ·  team project (ME 635, Group 11) A 4-robot warehouse fleet — dynamic task allocation, A* planning, VFH avoidance and token-based corridor right-of-way — with a 2D decision engine driving a physics-based 3D Simscape visualization. 12/12 tasks completed, zero collisions across 77 priority-yield events, no deadlocks. My role: designed the dynamic task allocator and the token-based FIFO corridor right-of-way protocol, built the entire 3D Simscape Multibody model from scratch, and created the 2D→3D data pipeline (pose/FSM logging, differential-drive wheel-angle derivation, frame transforms). (A* planning, VFH and analytics were teammates' contributions.)

Autonomous Object Sorting Using a Mobile Robot  ·  MATLAB / Simulink  ·  team project (ME 598) An autonomous mobile robot that navigates to objects and sorts them into predefined zones. 100% sorting accuracy across all six test configurations. My role: designed and implemented the collision-avoidance control law (distance-sensor-driven saturated-ramp velocity commands) keeping the robot clear of walls during navigation and plowing; co-developed the path-planning roadmap with teammates within a state-machine architecture.

Autonomy on real hardware, backed by validation discipline.